Overview of safety and compliance for autonomous machines
A landscape survey and gap analysis of safety and compliance across autonomous vehicles, robotic arms, and humanoid robots, and why the deterministic safety paradigm breaks down as intelligence moves to the center of the system.
Autonomous machines are entering everyday life faster than the safety and compliance frameworks meant to govern them. This paper presents a landscape survey and gap analysis of safety and compliance for three classes of autonomous machine: autonomous vehicles, industrial and collaborative robotic arms, and humanoid robots.
For each machine type, it surveys the current state of regulation and methodology, with the United States and European Union regulatory contrast threaded throughout. The paper then synthesizes the survey into a four-category gap analysis covering:
- Coverage gaps (where standards do not exist)
- Adequacy gaps (where standards exist but were not designed for the systems they are now being applied to)
- Coherence gaps (where jurisdictions conflict)
- Methodological gaps (where engineering methods are incomplete)
The analysis is grounded in published standards, regulatory documents, peer-reviewed literature from 2022 through 2026, and consultation with 20 practicing safety engineers, researchers, and consultants across the United States, European Union, India, and Japan, working at autonomous vehicle developers, humanoid robotics companies, industrial robotics manufacturers, compliance consultancies, and university research groups.
The central finding is that the industry is moving from "safe machines" to "safe autonomous systems operating in complex environments," a transition that established safety frameworks were not designed to handle. The frameworks were built around machines whose safety could be enforced by a predictable control layer wrapping the AI; as AI becomes the core of autonomous machines operating in complex environments, this control layer can no longer meaningfully constrain a system whose value depends on the AI itself. This shift is supported by independent diagnoses from the consulted practitioners and is now reflected in standards development inside the standards community itself, including the forthcoming ISO/IEC TS 22440 from the committee that maintains IEC 61508.
What you're reading is the abstract.